Despite the success of nationwide cervical smear screening programs in many developed countries, cervical cancers remain one of the leading cancers of women worldwide.1 Squamous cell carcinoma (SCC) and adenocarcinoma (AC) are the 2 most common histologic types, accounting for about 70% and 20–30% of all cervical cancers, respectively.2
Human papilloma virus (HPV) infection with integration of the viral DNA into the host genome is a crucial step of carcinogenesis of cervical cancers. The E6 and E7 viral oncoproteins can bind to p53 and retinoblastoma gene products, respectively; interact with various cell cycle-related cellular proteins; and result in antitumor suppression or disruption of cell cycle checkpoints.3, 4, 5, 6 Although high-risk type HPVs are highly prevalent among cervical cancers,7, 8 only a small percentage of high-risk HPV-infected cases progress to malignancy, which are usually after a long latency period.9, 10 The rates of progression to in situ or invasive cervical carcinoma vary considerably among HPV-positive women.11 Factors accounting for the long latency periods and genetic/epigenetic events governing the pace of progression have not been fully identified.
Loss of heterozygosity is commonly found in cervical cancers at chromosome 3p, 6q, 11q and 17q.12, 13, 14, 15, 16, 17 3p21 and 3p14 are 2 of the most frequent deletion regions, implying that cervical cancer-related tumor suppressor genes may be located at these sites.
FHIT (fragile histidine triad) located at 3p14.2 has been implicated as a possible tumor suppressor gene. Interestingly, FHIT spans a fragile site FRA3B, which is a frequent region of HPV integration.18 However, no significant differences in the prevalences of aberrant or truncated mRNA transcripts between cervical cancers and normal tissues were found.17, 19 Therefore, the role of FHIT gene in cervical carcinogenesis remains uncertain.
Recently, increasing evidence has been gathered to support that RASSF1A located at 3p21.3 is a possible tumor suppressor gene. RASSF1A inactivation via promoter hypermethylation is commonly found in several types of human cancer cell lines as well as primary human cancer tissue samples.20, 21, 22, 23, 24, 25, 26, 27, 28 Although hypermethylation of RASSF1A is frequent, true loss-of-function mutations are only rarely detected. Moreover, hypermethylation of RASSF1A has been proven to result in silencing of the gene: As hypermethylation was highly correlated with loss of gene expression, demethylation by 5′Aza-2′-deoxycytidine treatment resulted in restoration of gene expression.20, 21, 22, 23, 24, 25, 26, 27, 28 The tumor suppressor effects of RASSF1A has also been demonstrated by transfection of wild-type RASSF1A gene to hypermethylated lung cancers, which resulted in inhibition of tumor growth.21
In our study, we examined whether alterations of RASSF1A gene also occur in cervical cancers. Both SCC and AC were included. The results were analyzed with respect to the status of LOH at 3p21, HPV genotype, grading, staging and lymph node metastasis.
MATERIAL AND METHODS
Cervical cancer cell lines
Four human cervical cancer cell lines (C-33A, Hela, SiHa and ME-180) purchased from American Type Culture Collection (ATCC) were used for the study.
A total of 50 cases of cervical cancer (33 SCC, 17 AC), including cases of grades 1–3, and FIGO stages IA–IVA, and 11 normal cervical epithelia of Chinese females were included in our study. The specimens were retrieved from the tissue bank of the Department of Anatomical and Cellular Pathology of Prince of Wales Hospital of Hong Kong. The clinicopathologic information including age of patient at diagnosis, histologic type, tumor grade and stage (according to the International Federation of Gynecologists and Obstetricians [FIGO] staging system) as well as lymph node status were retrieved from the pathologic reports and case records. The cancer samples included hysterectomy, loop electrosurgical excisional procedure (LEEP) and cervical biopsy specimens. The age-matched samples of normal cervical epithelia were selected from hysterectomies performed for benign conditions. The histologic diagnoses of all cases were reviewed and reconfirmed by a pathologist (M.Y.Y.). Manual microdissection was performed in a way as described previously29 to isolate cells of interest for DNA extraction from formalin-fixed, paraffin-embedded tissue blocks. For each case of carcinoma, a corresponding normal tissue was used as control for LOH study.
Methylation-specific PCR (MSP)
The methylation status of the RASSF1A in the cervical cancer cell lines, primary cancers and normal cervical epithelia were examined by the MSP analysis as described previously.24 Genomic DNA was modified by bisulfite treatment using CpGenome DNA Modification Kit (Intergen, Purchase, NY) according to the manufacturer's instructions. The modified DNA was used as a template for MSP analysis. The primer sequences specific for methylated and unmethylated sequences were described previously.24 The PCR reactions were set up in a 25 μl reaction volume containing 1 × PCR buffer, 50–100 ng modified DNA, 5 pmole of each primer, 1.25 mM of each dNTP, 2 mM MgCl2 and 1 unit of AmpliTaq Gold polymerase (Applied Biosystems, Foster City, CA). The mixture was heated for 12 min at 95°C, followed by 40 cycles of denaturation at 95°C for 1 min, annealing at 60°C for 1 min and polymerization at 72°C for 1 min, followed by 5 min extension at 72°C on the GeneAmp PCR system 9600 (Applied Biosystems). Fifteen microliters of PCR products were loaded onto a 10% nondenaturing polyacrylamide gel, stained with ethidium bromide and visualized under UV illumination. In vitro hypermethylated DNA (IVD) (Intergen) was used as positive control for hypermethylation, and water was used as negative control.
Mutational analysis by single-strand conformation polymorphism (SSCP) and direct sequencing
DNA samples from primary cervical cancers were subjected to SSCP analysis for screening of RASSF1A gene. All 6 exons of RASSF1A gene were examined using 10 primer pairs as described previously.24 The PCR was carried out in 5 μl of reaction mixture containing 20–100 ng DNA, 0.25 pmole of each 32-P labeled primer, 12.5 mM of each dNTP, 2 mM MgCl2 and 0.25 unit of AmpliTaq Gold polymerase (Applied Biosystems) in 1 × PCR buffer. The mixture was heated for 12 min at 95°C, followed by 40 cycles of denaturation at 95°C for 1 min, annealed at 55–65°C for 1 min according to the recommended annealing temperature and underwent polymerization at 72°C for 1 min, followed by 5 min extension at 72°C on the GeneAmp PCR system 9600 (Applied Biosystems). Formamide dye of 45 μl was added to the PCR product to stop the reaction and aid electrophoresis. The mixtures of PCR product and formamide dye were denatured in boiling water for 10 min and chilled on ice. Two microliters of the product were immediately loaded onto a 6% nondenaturing polyacrylamide gel. Electrophoresis was carried out in 0.5 × TBE at 30W at 4°C for 6 hr or alternatively at 2W either at room temperature or 4°C overnight. After electrophoresis, the gel was completely dried and exposed to an X-ray film. The shifted bands on SSCP gels were isolated to elute DNA for sequencing analysis. The eluted DNA was reamplified and the PCR product was purified and subjected to direct sequencing using ABI PRISM DyeDeoxy Terminator Cycle Sequencing Kit and ABI 377 DNA Sequencer (Applied Biosystems).
The expression of RASSF1A in cervical cancer cell lines was examined by RT-PCR analysis. Total RNA from cell line samples was extracted by TriZOL reagent (Life Technologies, Rockville, MD). Seminested RT-PCR was performed using specific primers as described previously24 and listed in Table I. β-actin was used as internal control for RT-PCR.
To determine whether RASSF1A expression can be restored by demethylating agent, cervical cancer cell lines were subjected to 5′-Aza-2′-deoxycytidine treatment as previously outlined.30 Briefly, 5′Aza-2′-deoxycytidine (Sigma Chemical, St. Louis, MO) was added to the culture medium at the final concentration of 5 μM. After a 24 hr incubation with drug, the cells were rinsed with PBS, fresh culture medium was added and the cells returned to the incubator for 3 more days. On day 4, the cells were harvested for analysis of the methylation status of the RASSF1A.
Loss of heterozygosity (LOH)
DNA from the microdissected primary cancers and the paired normal controls were used for LOH analysis. Three polymorphic microsatellite markers (D3S1478, D3S1578 and D3S1076) encompassing chromosomal regions of 3p21 were employed. Fluorescent dye-labeled primer pairs for these markers were custom synthesized by Research Genetics (Huntsville, AL). The primer sequences, annealing temperature and estimated product size for each primer pair were available in Genome Data Base. Each PCR reaction mixture contained 50 ng of DNA, 1 × PCR buffer, 0.33 μM each primer, 2.5 mM MgCl2, 250 μM each deoxynucleotide triphosphate and 0.6 unit AmpliTaq Gold polymerase in a total volume of 10 μl. Amplified PCR products for multiple loci were pooled, electrophoresed on an ABI PRISM 377 automated DNA sequencer (Applied Biosystems) and analyzed with Genescan 2.1 software (Applied Biosystems). LOH was scored as described in our previous study.24
Human papilloma virus (HPV) detection and typing
The HPV status of primary cervical cancers were assessed by type-specific PCRs targeting the E7 regions of HPV-16, -18, -31, -33, -52 and -58, which are the most common oncogenic HPV types found in cervical cancers of Hong Kong Chinese women.8 We used the type-specific PCRs targeting E7 regions of HPVs because they are more sensitive for detection of the short DNA fragments archived from formalin-fixed, paraffin-embedded tissue samples as compared to the more conventional method that used the PCR assays such as MY09/11 targeting longer DNA fragments (450 base pairs) within the HPV L1 open reading frame (ORF).31
χ2 or Fisher exact test was used to study the association between categorical parameters including methylation status of RASSF1A, LOH at 3p21, HPV genotyping and clinicopathologic parameters. Statistical analyses were performed with the software system of SPSS version 10.1. A 2-tailed p-value <0.05 was regarded as statistically significant.
The results of status of methylation of RASSF1A, LOH at 3p21, HPV and clinical pathologic data are listed in Table II.
Table II. Clinical Parameters, HPV Status, 3p21-LOH, Hypermethylation and Mutation Analysis of RASSF1A in Cervical Cancers
LN metastasis: Yes, positive for ≥1 lymph node metastasis at pelvic or para-aortic regions; No, negative for lymph node metastasis; NA, information not available. L, LOH observed in ≥1 of the 3 markers (D3S1478, D3S1578, D3S1076); R, no LOH; NI, the case was not informative for the markers. M, hypermethylated; U, unmethylated; P, polymorphism with sequence change at codon 133.
HPV status of cell lines were obtained from American Type Culture Collection.
The ME-180 cell line showed more homology to HPV-39 than HPV-18.
The medium age of primary cancer patients was 44 years (range 27–80). The tumor grade and stage were determined according to the FIGO staging system. Twelve percent (6 of 50) of the cases belonged to histologic grade 1, 72% (36 of 50) grade 2 and 16% (8 of 50) grade 3. Thirty-five of 50 cancer patients (70%) were classified as stage I, 8 cases (16%) as stage II, 2 cases (4%) as stage III and 4 cases (8%) as stage IV. Eighteen percent (9 of 50) of the cancer patients were positive for pelvic or para-aortic regional lymph node metastasis. For normal controls, the medium age was 45 years, and the age range was 39–54 years.
All primary SCC cases were positive for one of the oncogenic HPVs: HPV-16 (26/33, 79%); HPV-18 (2/33, 6%), HPV-58 (4/33, 12%) or HPV-52 (1/33, 3%). (Table 2). The worldwide uncommon genotype HPV-58 was found to be the second most common HPV type in our SCC cases. This result concurred with our earlier study on Chinese females of the Hong Kong population.8 The HPV-positive rate for AC was 88% (15/17), whereby 53% (9/17) were HPV-16 and 35% (6/17) were HPV-18. As reported previously, HPV-18 were relatively more common in AC than in SCC.
LOH at 3p21
To determine the frequency of allelic deletion of RASSF1A locus in cervical cancer, we performed LOH analysis using 3 microsatellite markers flanking 3p21.3–21.1 (D3S1478, D3S1578 and D3S1076). Thirty-one of the 33 cases of SCC and all 17 cases of AC were informative for at least 1 of the markers. LOH at 3p21 was frequent in cervical cancer with 55% (17/31 informative cases) of SCC and 41% (7/17 informative cases) of AC showing loss at 1 or more loci. The overall frequency for 3p21 LOH in cervical cancer was 50%. Table II and Figure 1 show the LOH results. LOH at 3p21 was not correlated with patient age, HPV typing, tumor grade, FIGO stage or lymph node metastasis.
Mutation of RASSF1A
In 2 of 50 primary cancer cases (SCC-11 and SCC-25), identical mobility shifts were found in exon 3 of both cases, corresponding to a sequence change from GCT to TCT at codon 133 confirmed by DNA sequencing. This resulted in an amino acid substitution (alanine to serine). The same changes were identified in the paired normal tissues of these 2 cases, indicating that this is a germline polymorphism rather than a somatic mutation. However, no somatic mutation was detected in all primary cancer samples or cell lines (Table II and Fig. 2).
Hypermethylation and loss of expression of RASSF1A
Methylation status of the promoter region of RASSF1A gene was studied on 4 cervical cancer cell lines, 50 cases of microdissected cervical cancers and 11 normal cervical epithelia samples using methylation-specific PCR (MSP). Table II and Figure 3 show the results of MSP analysis. Complete methylation of RASSF1A was observed in 1 of the 4 cell lines, i.e., C-33A (Fig. 3), which was associated with loss of expression as demonstrated by RT-PCR. Demethylation treatment resulted in restoration of mRNA expression, indicating that methylation of RASSF1A was related to gene silencing. Hypermethylation of RASSF1A was present in approximately one-third of cases of SCC (10/33; 30.3%) and in approximately one-tenth of cases of AC (2/17; 12%) (Table II). The difference in methylation frequencies between SCC and AC was not statistically significant. Promoter hypermethylation was not detected in all normal controls. RASSF1A hypermethylation was not found to be correlated with patient age, HPV typing, tumor grade, FIGO stage or lymph node metastasis (see Table II).
Concomitant hypermethylation of RASSF1A and loss of heterozygosity at 3p21
Concomitant hypermethylation of RASSF1A and 3p21-LOH was found in a total of 8 cervical cancer cases, including 7 SCC and 1 AC. Seventy percent (7/10) of SCC cases with RASSF1A methylation demonstrated 3p21 allelic loss. Of 23 SCC with no RASSF1A methylation, 10 (43%) had allelic loss at 3p21. Conversely, 41% (7/17) among the group of SCC with 3p21 allelic loss showed RASSF1A methylation, whereas only 21% (3/14) of SCC with retention of 3p21 demonstrated RASSF1A methylation. For AC, only 2 of 17 cases (12%) were hypermethylated, with 1 of them showing both LOH at 3p21 and hypermethylation of RASSF1A (Table II).
3p21 is one of the most common regions for allelic losses in cervical cancers.12, 15, 17 We found LOH at 3p21 in 55% of cervical SCC and 41% of cervical AC, respectively, which is consistent with previously published data.13, 14, 16 The high frequency of allelic deletions suggests that these may be the sites resided with tumor suppressor genes. However, no known tumor suppressor gene for cervical cancers has been identified at 3p21.
RASSF1, a novel Ras effector gene located at 3p21.3, has recently gained many researchers' interest for its potential role as a candidate tumor suppressor gene. Ras possesses both oncogenic and anti-oncogenic properties.32 It requires other oncogenes, such as c-myc, to effect its full transforming activity.33 Although the mechanisms that determine Ras toward the oncogenic vs. the anti-oncogenic properties are not fully understood, some Ras effectors including Nore 1 and RASSF1 have been demonstrated to possess tumor suppression effects.34, 35 Five splice variants of RASSF1A to E have been reported. Vos et al.34 found that RASSF1C was able to bind to Ras in a GTP-dependent manner and transfection of RASSF1C to HIH 3T3 cells inhibit cell growth by inducing apoptotic-type cell death, which was p53 independent but ras dependent. In contrast, Avruch's group36 was unable to find GTP-dependent Ras binding abilities in RASSF1A, B or C. However, they showed that RASSF1A, but not B or C, homodimerizes or heterodimerizes Nore 1 and might thereby mediate the tumor suppressor effects of Ras through the interaction with Nore 1. Nore 1 was found to be able to form a complex with MST1 and acted as a novel Ras effector unit that mediated the apoptotic effect of Ki-RAS G12V.35
Although RASSF1C was demonstrated to induce Ras-dependent apoptosis in vivo and in vitro,34 no genetic or epigenetic changes of RASSF1C nor any underexpression of the gene were found in various human cancer cell lines or primary human cancers studied.21, 23, 26, 27, 28 On the other hand, frequent RASSF1A inactivation were detected in several major human cancers, notably in those cancers also known to show common allelic deletions at 3p21 regions, such as small cell lung carcinoma, mammary carcinoma, renal clear cell carcinoma, urinary bladder carcinoma and nasopharyngeal carcinoma.20, 21, 22, 23, 24, 25, 26, 27, 28 Hypermethylation of the promoter region of RASSF1A and loss of heterozygosity at 3p21 regions appeared to be the 2 most common mechanisms that lead to the gene silencing, whereas mutations were only rarely reported.20, 21, 22, 23, 24, 25, 26, 27, 28 It is still not fully understood how RASSF1A acts to suppress tumor growth. More recently, Shivakumar et al. found RASSF1A was able to inhibit tumor growth by blocking cyclin D1 accumulation and G1/S-phase progression of cell cycle.37
Our group addressed whether RASSF1A inactivation also occurs in cervical cancers. Dammann et al.21 had found loss of expression of RASSF1A by Northern blot analysis in the cervical cancer cell line C-33A whereas Hela cell line retained the expression. By RT-PCR analysis, we found similar results as Dammann's group. In addition, we also demonstrated complete methylation of RASSF1A in the cell line C-33A that related to gene silencing. For study of primary cervical cancers, we found RASSF1A hypermethylation in 30% of SCC and in 12% of AC. This indicated that RASSF1A was involved in at least a subset of cervical cancers and is indeed rather frequent in SCC, which is the most predominant histologic type of cervical cancers. Our result contrasted with another group, Agathanggelou et al., who studied 15 primary SCC and 7 AC and revealed no RASSF1A hypermethylation in any of their cases.20 The apparent discrepancy from our results may be related to their smaller sample size, and the possible ethnic or geographic differences cannot be excluded.
Somatic mutation of RASS1A is rarely reported in human cancer. Likewise, no somatic mutations of RASSF1A were found in all of our primary cancer samples or the 4 cervical cancer cell lines studied. However, in 2 of our primary cancer cases, a single nucleotide germline polymorphism at codon 133 of exon 3 (alanine-serine) was detected. This polymorphism has also been found in a nasopharyngeal cancer case by our group24 and in lung and breast cancer cell lines by Shivakumar et al.37 Interestingly, this rare allele has been shown by Shivakumar's group to encode a RASSF1A protein with impaired growth-inhibiting functions.37
Concomitant hypermethylation of RASSF1A and 3p21-LOH were frequently found in our primary cervical cancer samples. Approximately 67% (8 of 12) of hypermethylated primary cancers show concomitant 3p21-LOH. These findings supported that hypermethylation of RASSF1A and allelic loss at 3p21 were the 2 major mechanisms that resulted in inactivation of this putative tumor suppressor gene.
On the other hand, among the cervical cancer cases with 3p21-LOH, about half of them were not found to have any epigenetic or genetic changes with RASSF1A. This implied that other unknown tumor suppressor gene(s) located around this site are yet to be explored.
We have also examined the relation of hypermethylation of RASSF1A and 3p21-LOH status with HPV typing and clinicopathologic features. We found that all cases of SCC and 88.2% of AC were positive for the oncogenic HPV DNA (HPV-16, -18, -52 or -58). The results were consistent with the current data.7, 8 No correlations were found between hypermethylation of RASSF1A 3p21-LOH with HPV typing or with any of the clinicopathologic parameters. The lack of correlation with tumor stage implied that this gene may be involved in the early development of cervical cancer.
In summary, epigenetic inactivation of the RASSF1A gene via hypermethylation occurs in at least a subset of cervical cancers. Concomitant LOH at 3p21 is a common finding. The results supported that RASSF1A may be one of the cervical cancer-related tumor suppressor genes located at 3p21.
The authors thank Ms. L. Lin and Ms. H.Y.M. Wong for their technical support.